Reinforcement Learning-Based Energy Management Control Strategy of Hybrid Electric Vehicles

This article is aimed at developing a control strategy based on the Q-learning algorithm for HEVs. The Q-learning algorithm deals with high-dimensional state space problems, and the agent will have a “dimension disaster” problem during the training process. Then a control strategy based on the Deep Q Network (DQN) algorithm is introduced. Since DQN can only output discrete actions, in order to achieve continuous action control, an optimized control strategy based on the Deep Deterministic Policy Gradient (DDPG) algorithm is proposed. Simulation results show that compared with Q-learning and DQN algorithms, the DDPG algorithm converges faster, and the training process is more robust. Besides, the energy optimization control strategy based on the DDPG algorithm can better control the energy of HEVs.